# 1 - 2 - 3
Welcome everyone!
Our project is focused on how to use convolutional neuronal networks to detect diseases on cocoa fruit. Our project has real-world applications in our country to support our local farmers. Thank you for your attention and I gonna continue to this presentation.

# 4
Well, despite the enormous effort made by the world to reduce plant loss and food security, several references confirm that more than 20% of crop losses in the global scenario are due to plant diseases. This problem has worsened in the last decade due to the impact of pollution and climate change. With the recent development of various agricultural technologies, farmers opt for plant disease databases or consult local pathologists via telephones, instead of the classical procedure of sending plants to the diagnostic laboratory to propose the appropriate treatment.

# 5 - 6
Regarding plant disease detection, the goal of our project is on detecting potential threats to cocoa crops using a photograph that will be processed and analyzed using deep learning techniques. Deep learning is a powerful tool in image classification and has shown promising results in plant disease detection. Unlike traditional methods that use feature extraction algorithms such as SIFT, SURF, PCA, and LDA, deep learning is the preferred method for computer vision tasks. This technology offers a promising solution to ensure food security and sustainability in the future.

# 8
Ecuador is the leading cocoa producer in Latin America and the fifth largest in the world, although the lack of technology and resistance to pests and diseases are limiting factors.
Cocoa CCN-51 is a variety of cocoa originating from Ecuador and developed in the 1960s. It is known for its adaptability to different climatic zones, high productivity, and resistance to pests and diseases. However these characteristics are not enough to avoid the presence of diseases in the cocoa fruit.

Regarding the Cocoa pests and diseases, we have the following:
- Cocoa fly. Caused by the Monalonion dissimulatum bug, exclusive of cocoa, the insects feed on the shoots when young, and when they reach adulthood, they feed on the pods, causing pustules or circular wounds in the apical half of the fruit.
- Bull's horn. Caused by the sucking insect Hoplophorion pertusa, which in its adult stage feeds on the sap of the shoots and young branches, sucking the juices from the plant with its stylet. Excessive shade in the cocoa plantation predisposes to a greater attack by the pest.

# 9
Cocoa Diseases
- Moniliasis. Caused by the fungus Moniliophthera roreri, it affects the fruits, having variable symptoms according to the age of the fruit, as the infection progresses, a spot with white cottony tissue appears, this tissue turns grey due to the appearance of spores or seeds, ending with the mummification and deformation of the fruit.
- Witches' broom. Disease caused by the fungus Crinipellis pernicosa, it causes an abnormal sprouting at the level of both terminal and auxiliary buds, presenting a concentration of branches from a single point known as broom, in the affected floral cushions the flowers remain attached to this for a longer time than normal, developing unfertilised ovules, if the fruits are attacked it produces malformations similar to those caused by moniliasis [23].

# 11
About the metodology that we used.
Design Thinking and Scrum are two methodologies that can be combined to create an efficient and effective product development process. Design Thinking is used to understand user needs and desires, while Scrum is used for iterative product development and continuous delivery. This combination allows for user-centered solutions that can be delivered quickly and adapt to changing needs.

# 12
The selected architecture is based on the C4 Model where the specific model for the development of the API has been considered in its first iteration, this component will implement an onion architecture that will allow it to be tolerant to changes. The respective components for the web and mobile clients are external to the implementation and depends heavily on the API.

# 14
We define 4 phases in the project development

# 15
Web technologies such as Angular, FastAPI, SQLite, and Alembic were used to create a web platform that is both easy to build and robust. The idea was to develop a user-friendly web platform with a strong foundation.

# 16
Flutter, a mobile technology, was used in this project due to its ease and speed in developing interfaces that work independently of the operating system. We wanted to create a mobile platform that is user-friendly and can be developed quickly.

# 17
In this phase, several technologies were used, including:
- First we need to generate an image dataset, we gather around 5000 images from differently & nearly locations.
- We used LabelImg as a tool to label images in the dataset. The idea was to develop a system that uses cutting-edge technologies to accurately detect objects in images.
- The next step was to use the Faster R-CNN algorythm which internally uses another neural network called Region Proposal Network (RPN) that allowed us to generate proposals for regions of interest in the image.
- And the last tool was TensorFlow & Tensorflow Lite, the first one allowed us to work with the design that we had proposed and the second one which allowed us to use a model generated in our mobile app.

# 18
And then in the testing process, 100 external images were used to evaluate the model's performance. The model had to be adjusted several times to obtain results that satisfy the research goals. In this phase one of the main challenges encountered was adjusting the subnetworks for the regions of interest, which were trained separately and did not initially produce satisfactory results.

# 19
Azure cloud was used to centralize data collected from the application in one place. By using Docker and the services of Container Registry and Web App for Containers, any additional configuration was avoided and software updates were easily managed as the project progressed. The idea was to simplify the process of managing and updating the software while ensuring that all data was securely stored in one location.

# 21 - 22
As you can see, during the application usage process, when a user takes a photo, it is first stored on the server and then analyzed using a pre-trained and exported model within the same application. The idea was to ensure that the analysis process was seamless for the user while ensuring that the data was securely stored on the server.

# 23
The application has a success rate of 80 to 99.5%. These results were obtained after subjecting the application to different environments and climate changes that could affect the quality of the photographs.
In addition, a comparison was made between the traditional methods in the area and the application, obtaining the following results.

the traditional method had a 30 to 50 failure percent and 50 hit percent
The Laboratory analysis had a 0.1 to 1 failure percent and 99 hit percent
and finally the Application had a 0.4 to 20 failure percent and 80 to 99 hit percent

# 25
The creation of tools and means of technology that allow a faster reach in the prevention of diseases in the production fields, is an alternative that will allow the growth and improvement of the product, in addition to a quick prevention eliminates part of the chemicals used for the cure of these pests, this improves both the quality of the product and the shelf life of the plant, maintaining the taste of the same.
The long-term goal of this project is to expand the application to cover a wider area beyond its current scope. Specifically, we aim to extend the reach to a larger region that could benefit from this technology.